Modifying inMoov for Robotis Dynamixel Servos

[sla...@gmail.com]

Hello Gael,

My research group at the University of Alberta is developing a training tool to help upper limb amputees learn how to control their muscle signals and use myoelectric technology in advance of their myoelectric prosthesis fitting.

You can see here one of our previous papers on an early version of our myoelectric training tool:
http://www.ncbi.nlm.nih.gov/pubmed/22383905

We have in the past used crustcrawler's AX-12/AX-18 smart arm as our robotic arm platform, but are currently looking to move towards something more anthropomorphic. We are consequently very interested in your inMoov project especially in the hand and the recently released prosthetic finger assembly.

I am interested in contributing to the project in the following possible ways:

1) Adapting the hand and potentially lower/upper arm to function with Robotis' Dynamixel servos, which are quite a bit more versatile in research or even hobby applications since they provide position and velocity control as well as feedback on position, velocity, load, temperature, and voltage while still being cost-effective compared to industrial servos. Several API's and drivers are available and are supported on all the major platforms including windows/linux/mac and embedded options such as Arduino. Here is a link to the Robotis website for more information:
http://www.robotis.com/xe/dynamixel_en

2) Experimenting with using PLA as the build material. We have access to both a replicator 2 and a replicator 2x, so it is possible for us to try both materials and compare how they work with the inMoov. Currently, we are favoring the replicator 2 and the PLA option as we are getting more consistent results and a higher build quality out of it, so it may be interesting to try and tweak some of the inMoov parts to fit better with PLA.

3) Integrating sensors into the hand for force feedback applications.

In order to facilitate doing these things I was wondering if there is someway I could get a hold of the original blender files for the hand, lower arm, and upper arm. I mostly use solidworks for 3d modelling, so I was going to see if there is someway to import the blender files and more easily modify/adapt the parts. I have tried importing the STL's but the process does not work very well with solidworks and the model comes in with many facets instead of smooth curves. If importing the STL's is the only option it will probably be easier to remodel them, which could be very time consuming. In any case let me know as I'm excited to start adapting and printing some parts.

Thanks,

Rory Dawson
 

[Ben Barbour]

I would be interested to see what you can come up with as well.  I have been editing the STLs to work with PLA for a while now.  The primary issue with PLA has been that it is so rigid that there is no room for error for interlocking parts.  Many of the exiting parts are designed to be a snug fit with ABS's flexibility.  I've had some success but lately I've been on hold because of a servo issue with my Replicator 2.

[Joel Hackett]

Ben I had a problem with my replicator 2, it was the y axes stepper motor belt was over tight, it was stalling out / slipping.

[gael langevin]

Hi all,

Dynamixel servos are definitely much better servos than RC servos. When I started the hand someone suggested to use the Dynamixel but the price made me look somewhere else. I think if we can have both options it is of course a very nice opportunity.

The Dynamixel have a lot more features and advantages to give. I'm wondering though about the micro-controller, can we still use Arduino?

I had chosen Arduino to stay Open source as I also choose MyRobotLab, to avoid to get tight up by some companies.

The issue about importing STLs has been a recurrent question since the beginning. Using Blender is a non formal way of doing mechanical parts, I agree. But I found my way because of Blender which is Open source, Solidworks is expensive unless you can use it in a University or a Company. Dassault Systèmes SolidWorks Corp.

The way I'm using Blender is even less conventional than you would think, unfortunately.

I export my parts in STL and re-import them in another Blender file as STLs, so what I have in my original Blender files are STLs, same as what you can find for downloads.

There is a way to generate CAD files from STLs but I actually never tried it myself:
http://robosavvy.com/forum/viewtopic.php?p=34865

2013/7/18 Sebastian Schocke <hunt...@gmail.com>

Wow, I am very impressed with those Dynamixel servo's. A bit on the expensive side for messing around with, but definitely something to keep in mind should one ever go for a more reliably useful robot at the end of the day. Something I would definitely look at if the parts can be redesigned to use them. What do you say Gael?

--
You received this message because you are subscribed to the Google Groups "InMoov" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com.
To post to this group, send email to inm...@googlegroups.com.
Visit this group at http://groups.google.com/group/inmoov.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[AVRnj]

Hey everyone,

New here, but I have been checking out inMoov for the past few days.

First off, let me just say I am in awe of what you have done here Gael, and making this available the way you do is incredible.

I am a huge fan of the Dynamixel servos, and ironically I just finished this am modifying the simple finger starter to use with some Dynamixel's I had laying around.

I took the STL's and mucked around with them a bit in openSCAD which is what I am most familiar with, and printed it out this am and tested it out.

It worked well, although my STL's are a bit sloppy and the fit was not perfect. The pulley I made worked well, but the bracket for the servo itself was a bit off.

Yes, Dynamixels can most definitely be used with Arduino. There are a few libraries out there that support them. Dynamixels work using Serial, by sending packets of instructions to the Dynamixel itself.

I have a library that I started with that I have improved dramatically, as I am a software developer.

Many things are incredible about Dynamixel, but the one thing that lends itself so nicely to this project is you can daisy chain them, so only one servo needs to connect to your Arduino, and the rest can be daisy chained. This makes much cleaner wire distribution.

The other thing is the newer ones all work off of Serial communication, and in particular TTL half duplex. So what this means is you can interchange different series of Dynamixels on the same chain. So say you want to use something like an AX18 for a finger, but say an MX106 for the bicep, you can do so. You can also swap out an AX for an MX later without changing any code at all, although you do get better code functionality with the MX series than you do for AX.

IMO Dynamixels are so far superior to hobby servos in every way. You have so much control over them, torque, speed, exact angle, you can ask it what angle it's at etc.

I actually have been thinking of doing an inMoov using Dyanmixels for the past few days now, and it would be great if others were interested as well, as there will need to be a lot of modifcations to the servo holders to make this work.

Anyway, sorry for the long first post, hope this was somewhat helpful and I can contribute to this awesome project.

[gael langevin]

That's good news about the use of Arduino with Dynamixel. The Daisy chaining was something that I liked about them, not mentioning all the other features you are referring.

So you tried to swap the servo for finger starter with a AX18? They also have different actuator axe diameter, correct?

One problem you will have is space. Dynamixel servos are big, I was just looking at the AX12, the size is much bigger than a MG996r for the same torque.

Fitting five of these in the forarm doesn't only mean re-designing the brackets but a lot more I think.
Gee, five of these cost already 400 dollars...

AX12 is the smallest there is with this stall torque, correct?

2013/7/19 AVRnj <ey74...@gmail.com>

[AVRnj]

I am not sure about the actuator axle diameter in comparison to other servos to be honest.

Actually, I swapped the servo for finger starter with an AX12, which is the entry level Dynamixel.

I see what you mean about space, your right, those MG996r's are nice and small, and cheap compared to the Dynamixels!

Im not sure what country you are in, I am in the US, and I get 6 of them for $225, which is reasonable given their functionality.

Correct, the AX12 is the smallest Dynamixel made.

There is another brand of smart servo called HerkuleX which are a bit cheaper, but a similar idea, similar in size as well.

The other thing that is really good about the Dynamixels are that they have temperature sensors built in, and you can program them to stop if they get in the danger zone, saves a lot of servos. They will actually sound an alarm.

Here is a link to a quick video of the simple finger I printed today and using an AX12A, it's a bit rough and tight fitting so not as smooth as it should be, but I just wanted to do something quick to give it a shot.

https://dl.dropboxusercontent.com/u/42248323/ABS/IMG_1064.MOV

Sorry for the bad video quality, shot on my iPhone in bad lighting.

[gael langevin]

Thanks for the video!
 It actually doesn't look that big, they are larger than the RC servos and if you see how the servos are arranged in the forarm you will see that there is'nt much extra space, but I guess it should be possible.

I was just reading this when I received your new comment about the alarm for temperature sensing.
http://www.shervinemami.info/dynamixels.html

France actually sell them more expensive than the price you get in the US apparently. Although the price I found was a quick search.

[AVRnj]

Interesting, I had never seen that site before.

Looks like another library, I use a different one.

They are reasonable in size. I have only used a few hobby servos, HS-311's and they are comparable in size to those, but maybe not as small as the ones you are using.

Once I started using the Dynamixels, I found it hard to go back to hobby servos, but the price and size is a concern for this project, given the aesthetics you are achieving, which are remarkable, and also you have so many DOF that the price is a concern with Dynamixels.

I am still debating whether I want to do this with the Dynamixels or another project I have in mind. Perhaps a hybrid of the two.

[Samuel Presgraves]

I would LOVE this, I'm trying to build my inmoov with dynamixels, but I suck at editing 3d files.  I plan on using a variety of AX-12A, AX-18A, and MX-28T's.  Let me know if you make headway with this

[Stefan Eriksson]

I like the Dynamixel servos for those extra features but the price-tag is way to expensive :(
Im a poor worker and dosent have 20 000 dollars (or more) to spend on a robot for fun...

The Dynamixel Pro is whats needed for the legs and shoulder to cope with the forces without gears.

And is the structure itself capable to take that extra punishment?
Maybe a rebuild to strengthen InMoov is needed?

I have looked at Charli and his big brother Thor that uses Dynamixel servos and they are nice robots but not in a "hobbyist" price range :( 

For university that have the funds its good, so im looking forward to see your robot! :)


Den onsdagen den 17:e juli 2013 kl. 21:35:35 UTC+2 skrev salivadontist:

Hello Gael,

[Duane Mathes]

Have you guys made any progress with the Dynamixel servos? I have a bunch laying around at my work and I'd like to use them for the bicep and shoulder. Have you guys come up with a library for running the RS485 on the Dynamixels instead of the normal PWM of hobby servos? 

Duane

[Samuel Presgraves]

Come over to the MRL site, I'm working on a FPGA based design that's aimed at dynamixes for the primary servos. I should have the first revision available in the spring.